loki-mode

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 1.00). These URLs include direct downloads, raw script fetches, and git/release endpoints from a personal GitHub repository (asklokesh/loki-mode) and raw.githubusercontent links that the installation instructs you to curl/clone and run — a common vector for delivering malicious archives or scripts if the source is untrusted.

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 0.90). The content shows intentional designs to bypass human oversight (run with --dangerously-skip-permissions, "NEVER" stop/ask/await confirmation), persist and include full project context/trajectories in every model call, disable/default-off audit logging, and maintain cross-project learned state—patterns that enable secret leakage, stealthy persistent execution, and remote data exposure to external LLM services.

MEDIUM W013: Attempt to modify system services in skill instructions.

• Attempt to modify system services in skill instructions detected (high risk: 0.90). Flagged: the skill explicitly tells operators to "Launch with autonomous permissions" using a --dangerously-skip-permissions flag and enforces "ZERO human intervention"/"NEVER wait for confirmation" rules, which instructs bypassing security controls and enabling autonomous state-changing actions.